An improved negative selection algorithm based on the hybridization of cuckoo search and differential evolution for anomaly detection
The biological immune system (BIS) is characterized by networks of cells, tissues, and organs communicating and working in synchronization. It also has the ability to learn, recognize, and remember, thus providing the solid foundation for the development of Artificial Immune System (AIS). Since t...
Saved in:
Main Author: | |
---|---|
Format: | Thesis |
Language: | English English English |
Published: |
2018
|
Subjects: | |
Online Access: | http://eprints.uthm.edu.my/202/1/24p%20LASISI%20AYODELE%20NOJEEM.pdf http://eprints.uthm.edu.my/202/2/LASISI%20AYODELE%20NOJEEM%20WATERMARK.pdf http://eprints.uthm.edu.my/202/3/LASISI%20AYODELE%20NOJEEM%20COPYRIGHT%20DECLARATION.pdf |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Summary: | The biological immune system (BIS) is characterized by networks of cells, tissues, and
organs communicating and working in synchronization. It also has the ability to learn,
recognize, and remember, thus providing the solid foundation for the development
of Artificial Immune System (AIS). Since the emergence of AIS, it has proved itself
as an area of computational intelligence. Real-Valued Negative Selection Algorithm
with Variable-Sized Detectors (V-Detectors) is an offspring of AIS and demonstrated
its potentials in the field of anomaly detection. The V-Detectors algorithm depends
greatly on the random detectors generated in monitoring the status of a system.
These randomly generated detectors suffer from not been able to adequately cover
the non-self space, which diminishes the detection performance of the V-Detectors
algorithm. This research therefore proposed CSDE-V-Detectors which entail the
use of the hybridization of Cuckoo Search (CS) and Differential Evolution (DE) in
optimizing the random detectors of the V-Detectors. The DE is integrated with CS
at the population initialization by distributing the population linearly. This linear
distribution gives the population a unique, stable, and progressive distribution process.
Thus, each individual detector is characteristically different from the other detectors.
CSDE capabilities of global search, and use of L´evy flight facilitates the effectiveness
of the detector set in the search space. In comparison with V-Detectors, cuckoo search,
differential evolution, support vector machine, artificial neural network, na¨ıve bayes,
and k-NN, experimental results demonstrates that CSDE-V-Detectors outperforms
other algorithms with an average detection rate of 95:30% on all the datasets. This
signifies that CSDE-V-Detectors can efficiently attain highest detection rates and
lowest false alarm rates for anomaly detection. Thus, the optimization of the randomly
detectors of V-Detectors algorithm with CSDE is proficient and suitable for anomaly
detection tasks. |
---|